Use "take the best": choose on your single most informative cue

When choosing between options, identify the most diagnostic cue and use it — stop searching for more.

Why it works

Take the Best is a lexicographic heuristic: rank your cues by validity, check the best one first, and make your decision if it discriminates. Only consult further cues if the top one doesn’t discriminate. This ignores a great deal of information — and in uncertain real-world environments, it often outperforms regression models that use all the information, because each additional cue adds noise alongside signal.

How to do it

1. For a recurring decision, write down the three most predictive cues you’ve observed historically.
2. Order them by how often they’ve been right when they pointed in a direction.
3. When facing the decision, check the top cue first. If it clearly points one way, stop — act on it without consulting the others.

Evidence

Gigerenzer and colleagues showed that Take the Best outperformed multiple regression on real datasets in out-of-sample prediction across several domains, particularly when data was limited and the environment was uncertain. (observational)

Take the Best outperforms regression under specific conditions — noisy environment, limited data, correlated cues. In large, clean datasets with many valid independent predictors, regression typically wins.

Sources

• Gigerenzer & Goldstein (1996), Reasoning the fast and frugal way, Psychological Review

Common mistake

Using Take the Best in data-rich, stable environments where all cues carry independent signal — here, looking at more cues genuinely helps, and stopping at one is costly.